Blog

/

AI & trust

We don't train on your data. Here's what that means

Your conversations and docs never improve anyone else's model, ours included. A plain-language look at how we keep that promise.

Ravi Menon

Head of Trust & Security

5 min read

Article cover image

"We don't train on your data" is easy to say. Here is exactly what it covers, and how you can check it.

What stays yours

Your help docs, customer conversations and settings are used only to answer your customers. They are never added to a training set, shared between workspaces or sold.

How we enforce it

  • Each workspace has its own isolated index

  • Model providers are contracted for zero data retention

  • Every access to customer content is logged and reviewable

Enterprise customers can also choose US or EU hosting and set their own retention period.

Next article →

Measuring support quality beyond CSAT

Written by

Ravi Menon

Head of Trust & Security

Let Relay answer the easy ones.

Live on your site in 10 minutes. 14-day free trial, no card.

● 14-day free trial

Give your team their evenings back.

Join 2,400+ support teams that let Relay handle the repetitive questions.

Resolved just now

Refund issued for #4821

Maya · Parcelly

38s

Password reset link sent

Leo · Lumen

12s

Plan upgraded to Pro

Ines · Orbitly

54s

Identity verified, balance shared

Kofi · Vaultio

41s

Relay

AI support that sounds like you. Built for teams of 2 to 200.

Sales replies within 1 business day

Support tips, once a month.

Real playbooks from teams using Relay. No spam.

© 2026 Relay Inc. All rights reserved.

All systems operational

Relay

AI support that sounds like you. Built for teams of 2 to 200.

Sales replies within 1 business day

Support tips, once a month.

Real playbooks from teams using Relay. No spam.

© 2026 Relay Inc. All rights reserved.

All systems operational

Create a free website with Framer, the website builder loved by startups, designers and agencies.